Anomaly Detection in Internet of Things Based on Logs Using Machine Learning and Deep Learning Techniques
Huy‐Trung Nguyen, Viet Quoc Nguyen · 2023
Engineers (developers or operators) can comprehend the condition of the system and spot odd behaviors like malware attacks and system failures by using log data that records critical events and system status. However, in the 4.0 era, IoT devices are expected to explode in number, and a large amount of data is generated from IoT devices. If something goes wrong, engineers will spend a lot of time manually processing expansive sums of log information. Therefore, it is vital to develop automated methods for for log-based anomaly detection, machine learning, and deep learning applications. However, with IoT device log data, how compelling are straightforward deep learning and machine learning models, and which approach will be more reasonable? This work is for research and evaluation of machine learning and deep learning models with two actual log datasets. The machine learning algorithms like RF, kNN, XGBoost models are trained on two actual log datasets based on the log parsers. Ensemble classifier, XGBoost got the best results with Accuracy, precision, and F1-score best at 99.9%, 99.8%, and 99.9%, respectively. We expect that the discoveries of our think about will be very beneficial for both professionals and analysts seeking after this interesting field.